GELE Adjustment Computations (Least Squares) — Theory of Errors, Weights and Most Probable ValueDetailed Explanation
Detailed explanation of Theory of Errors, Weights and Most Probable Value for the GELE 2026. Full depth, full reasoning — exactly what you need when Professional Regulation Commission (PRC) — Board of Geodetic Engineering tests this chapter with applied or scenario-based questions in the GELE Adjustment Computations (Least Squares) subtest.
Exam context
For the Geodetic Engineer Licensure Examination, Professional Regulation Commission (PRC) — Board of Geodetic Engineering tests Adjustment Computations (Least Squares) under a "Core" label, with Theory of Errors, Weights and Most Probable Value in the 1st slot across 5 chapters. GELE candidates must clear the 70% weighted average, no sub-test below 50% cut on the 2026 paper, which draws about a meaningful share of Adjustment Computations (Least Squares) questions. Date to watch: September 2026.
Theory of Errors, Weights and Most Probable Value - Detailed Explanation
In geodetic engineering, no measurement is perfect. Every observation — whether a horizontal angle measured with a theodolite, a distance measured with a total station, or an elevation determined by differential leveling — contains some degree of error. The Theory of Errors provides the mathematical and statistical foundation for understanding, classifying, and managing these inevitable imperfections. The concept of Weights quantifies the relative reliability of each observation, while the Most Probable Value (MPV) gives us the single best estimate from a set of redundant measurements. Together, these three pillars form the theoretical backbone of Adjustment Computations using Least Squares — a core competency tested in the PRC Geodetic Engineer Licensure Examination. This chapter equips you with the precise vocabulary, formulas, and problem-solving skills needed to confidently answer board exam questions on this fundamental topic.
Concepts
Classification of Errors in Measurement
Every measurement deviates from the true value. This deviation is called an error. In adjustment computations, errors are classified into three distinct types: Blunders (Mistakes), Systematic Errors, and Random (Accidental) Errors. Understanding each type is critical because each demands a different response from the geodetic engineer. **1. Blunders (Mistakes)** Blunders are gross errors caused by human carelessness, equipment malfunction, or misreading. Examples include: transposing digits (recording 58.32 m instead of 53.82 m), misreading a leveling rod, or forgetting to level a total station. Blunders do NOT follow any statistical law and therefore CANNOT be adjusted. The correct approach is to DETECT and ELIMINATE them before any adjustment is performed. Detection methods include field checks, closure computations, and statistical outlier tests. **2. Systematic Errors** Systematic errors follow a definite pattern or law and always act in the same direction under the same conditions. They are caused by identifiable sources such as: instrument imperfection (e.g., index error of a theodolite), environmental conditions (e.g., refraction, temperature), or procedural methods. Because they follow a law, systematic errors can be MODELED and CORRECTED. For example, in EDM distance measurement, a scale correction is applied for atmospheric conditions. In leveling, a curvature and refraction correction is applied. Systematic errors CANNOT be eliminated by averaging more observations — in fact, more observations only accumulate the same bias. **3. Random (Accidental) Errors** Random errors are the small, unpredictable residuals that remain after blunders have been eliminated and systematic errors have been corrected. They arise from the human inability to make a perfect measurement — slight variation in pointing, reading, and leveling. Random errors obey the Normal (Gaussian) probability distribution and have these four fundamental properties: - They are equally likely to be positive or negative (compensating). - Small errors occur more frequently than large errors. - Very large errors are extremely rare. - In the limit of many observations, they tend to cancel. Because of these properties, random errors are HANDLED through statistical adjustment — specifically, the Method of Least Squares. The goal of adjustment is to find the set of values that minimizes the sum of the squares of the residuals: Σv² = minimum.
Examples
The small, sign-varying discrepancy of +2' is characteristic of random error. It is small, and its cause (slight variations in pointing and reading the theodolite) is unpredictable. This error is handled by distributing the correction (-2'/3 = -0°00'40'' per angle) — a form of least squares adjustment.
Scenario
A surveyor measures the interior angles of a triangle as 60°01', 59°58', and 60°03'. The angular closure error is +2'. Classify this error.
Solution
The sum of the interior angles is 60°01' + 59°58' + 60°03' = 180°02'. The true sum must be 180°00'. The closure error of +2' is a random error distributed among the three angle measurements.
The error follows a definite physical law (Snell's Law applied to EDM). It always acts in the same direction (either all measurements are too short or all too long). The correction must be computed and applied before adjustment.
Scenario
During EDM distance measurement, the temperature was 35°C but the instrument was set for a standard temperature of 15°C. Classify this error.
Solution
This is a systematic error due to an uncorrected atmospheric condition. The distance correction formula is: ΔD = D × (n_actual - n_standard), where n is the refractive index. The error consistently shortens or lengthens all measured distances in the same direction.
Applications
- Field quality control: Identifying and rejecting gross errors before office computations.
- Instrument calibration: Determining and correcting systematic index errors of theodolites and levels.
- Least squares adjustment of GPS baselines: Random errors in phase observations are adjusted by least squares.
- Cadastral surveys under PD 1529: Closure checks detect blunders in traverse surveys before titling.
- NAMRIA geodetic network adjustment: Random errors in angle and distance observations across the PPCS/UTM grid are adjusted by least squares.
Misconceptions
- WRONG: 'Averaging more measurements eliminates all errors.' CORRECT: Averaging reduces the effect of random errors but accumulates systematic errors.
- WRONG: 'A small error is always a random error.' CORRECT: Even a very small systematic error remains systematic and must be corrected, not adjusted.
- WRONG: 'Adjustment computations handle all three types of errors.' CORRECT: Adjustment only handles random errors. Blunders must be eliminated first, and systematic errors corrected first.
Related Concepts
- Normal Distribution of Random Errors
- Probability Theory in Measurements
- Outlier Detection (Data Snooping)
- Propagation of Errors
- Instrument Calibration and Corrections
Common Exam Questions
Example
Q: A leveling rod was read as 1.523 m but the actual reading was 1.253 m. What type of error is this? A: Blunder (transposition of digits 5 and 2).
Approach
Read the scenario carefully. Ask: Is it a large, obvious mistake? → Blunder. Does it consistently affect the measurement in one direction? → Systematic. Is it small, random, and sign-varying? → Random.
Question Type
Classification/Identification
Example
Q: Why does taking the average of multiple observations NOT eliminate systematic errors? A: Because systematic errors act consistently in the same direction; averaging only repeats the same bias.
Approach
Focus on which error type is eliminated, corrected, or adjusted, and why.
Question Type
Conceptual Understanding
Key Points To Remember
- Blunders: Eliminate before adjustment — they CANNOT be adjusted.
- Systematic errors: Model and CORRECT them — they follow a definite law.
- Random errors: ADJUST them using least squares — they follow the normal distribution.
- More observations reduce the effect of RANDOM errors (by √n), but ACCUMULATE systematic errors.
- The correct sequence is: Eliminate blunders → Correct systematic errors → Adjust random errors.
- Board exam favorite: 'Which type of error is handled by adjustment computations?' Answer: Random errors.
Most Probable Value (MPV) and Arithmetic Mean
The **Most Probable Value (MPV)** is the single best estimate of the true value of a measured quantity, derived from a set of redundant observations. 'Most probable' means it is the value most likely to be closest to the true value, based on the principle of least squares. **For equally reliable (equally weighted) direct observations**, the MPV is the **arithmetic mean**: $$\bar{x} = \frac{\sum x_i}{n} = \frac{x_1 + x_2 + \cdots + x_n}{n}$$ This result is not just a convention — it is derived from the principle of maximum likelihood applied to the normal distribution. The arithmetic mean is the value that minimizes the sum of squares of residuals (Σv² = minimum), which is the fundamental criterion of least squares. **Residuals** A residual (v) is the difference between an individual observation and the MPV: $$v_i = x_i - \bar{x}$$ Note: Residuals are NOT true errors. True errors are ε_i = x_i - X_true, but since the true value X_true is unknown, residuals are our best estimate of the errors. An important property of residuals from the arithmetic mean is that their sum is always zero: Σv_i = 0. This serves as a computational check. **Standard Deviation of a Single Observation** The standard deviation σ measures the spread or dispersion of individual observations about the mean: $$\sigma = \sqrt{\frac{\sum v_i^2}{n - 1}}$$ Note the denominator is (n − 1), not n. This is Bessel's correction — it gives an unbiased estimate of the population standard deviation when working with a sample of n observations. **Standard Deviation of the Mean** The mean of n observations is more reliable than any single observation. The standard deviation of the mean (also called the standard error of the mean) is: $$\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}}$$ This shows the powerful effect of redundancy: taking 4 times as many measurements halves the uncertainty of the mean; taking 100 times as many reduces it by a factor of 10. **Probable Error** The probable error (PE) is another measure of precision, related to σ by: $$PE = 0.6745\sigma$$ The probable error defines the range within which there is a 50% probability of finding the true error. The probable error of the mean is: PE_mean = 0.6745σ/√n.
Examples
The check Σv = 0 confirms the arithmetic. Using (n−1) = 4 in the denominator gives the unbiased sample standard deviation. The mean has a standard deviation of only 0.002 m — considerably better than the single-measurement σ of 0.004 m, demonstrating the benefit of redundant observations.
Scenario
Five repeated measurements of a horizontal distance using a calibrated steel tape give the following values (in meters): 125.342, 125.338, 125.345, 125.340, 125.335. Find: (a) the MPV, (b) the residuals, (c) the standard deviation of a single measurement, (d) the standard deviation of the mean.
Solution
(a) MPV = Σx/n = (125.342 + 125.338 + 125.345 + 125.340 + 125.335)/5 = 626.700/5 = 125.340 m (b) Residuals: v₁ = 125.342 − 125.340 = +0.002 m v₂ = 125.338 − 125.340 = −0.002 m v₃ = 125.345 − 125.340 = +0.005 m v₄ = 125.340 − 125.340 = 0.000 m v₅ = 125.335 − 125.340 = −0.005 m Check: Σv = +0.002 − 0.002 + 0.005 + 0.000 − 0.005 = 0.000 ✓ (c) Σv² = (0.002)² + (0.002)² + (0.005)² + (0)² + (0.005)² = 0.000004 + 0.000004 + 0.000025 + 0 + 0.000025 = 0.000058 m² σ = √(0.000058/4) = √(0.0000145) = 0.00381 m ≈ 0.004 m (d) σ_x̄ = σ/√n = 0.00381/√5 = 0.00381/2.2361 = 0.00170 m ≈ 0.002 m Final answer: MPV = 125.340 m ± 0.002 m
To triple the precision of the mean (reduce σ_x̄ by a factor of 3), you need 3² = 9 times as many observations. This illustrates the diminishing returns of adding more measurements — quadrupling observations only doubles precision.
Scenario
A single survey observation has σ = 0.015 m. How many observations are needed to achieve a mean standard deviation of 0.005 m?
Solution
σ_x̄ = σ/√n 0.005 = 0.015/√n √n = 0.015/0.005 = 3 n = 9 observations
Applications
- Geodetic control surveys: Computing the best estimate of a benchmark elevation from multiple leveling lines.
- GNSS positioning: Computing the MPV of repeated static GPS coordinates.
- Angle measurement with a transit: The mean of multiple direct/reverse readings is the MPV.
- Quality control in cadastral surveys under PD 1529: Checking if the standard deviation of closure meets acceptable limits.
- EDM calibration: Computing the MPV of repeated calibration baseline distances.
Misconceptions
- WRONG: 'The MPV is the median or mode of the observations.' CORRECT: For normal distribution of random errors, the MPV is the arithmetic mean.
- WRONG: 'Residuals and errors are the same thing.' CORRECT: Residuals are computed from the sample mean (v_i = x_i − x̄); true errors require the unknown true value (ε_i = x_i − X_true).
- WRONG: 'Use n in the denominator of σ.' CORRECT: Use (n−1) for sample standard deviation. Using n would underestimate σ.
- WRONG: 'Taking 100 measurements gives 100x better precision of the mean.' CORRECT: Precision improves only by √100 = 10 times.
Related Concepts
- Weighted Mean (for unequal weights)
- Least Squares Principle (Σv² = minimum)
- Normal Distribution and Probability
- Propagation of Random Errors
- Confidence Intervals
Common Exam Questions
Example
Q: Ten measurements of an angle give Σv² = 64 (arcseconds²). Find σ and σ_x̄. A: σ = √(64/9) = √7.111 = 2.67''; σ_x̄ = 2.67/√10 = 0.844''.
Approach
Step 1: Compute x̄ = Σx/n. Step 2: Compute residuals v_i = x_i − x̄ and check Σv = 0. Step 3: Compute Σv². Step 4: σ = √(Σv²/(n−1)). Step 5: σ_x̄ = σ/√n.
Question Type
Computation — Find MPV, σ, σ_x̄
Example
Q: If σ = 0.030 m, how many observations give σ_x̄ = 0.010 m? A: n = (0.030/0.010)² = 9.
Approach
Use σ_x̄ = σ/√n. Solve for n = (σ/σ_x̄)².
Question Type
Conceptual — Number of observations for target precision
Key Points To Remember
- MPV for equal-weight observations = Arithmetic Mean = Σx/n.
- Residual: v_i = x_i − x̄ (observation minus MPV). Always check: Σv_i = 0.
- Standard deviation of a single observation: σ = √(Σv²/(n−1)). Use (n−1) in the denominator — Bessel's correction.
- Standard deviation of the mean: σ_x̄ = σ/√n. Precision improves as √n, NOT as n.
- Probable error = 0.6745σ. Probable error of mean = 0.6745σ/√n.
- The MPV minimizes Σv² — this IS the least squares criterion.
- Board exam trap: Using n instead of (n−1) in the σ formula gives a wrong answer.
Weights and Weighted Mean
Not all observations are equally reliable. An angle measured once has more uncertainty than the same angle measured ten times. A leveling line running 1 km has higher quality than one running 10 km. To account for these differences in reliability, we assign a **weight** to each observation. **Definition of Weight** The weight of an observation is a dimensionless number expressing its reliability relative to the other observations. A higher weight means the observation is more trustworthy and should have more influence on the final result. Fundamentally, weight is **inversely proportional to variance**: $$w_i = \frac{c}{\sigma_i^2}$$ where c is any positive constant (often chosen so that one weight equals 1, or so that the smallest/simplest weight is a whole number). The critical insight: weight depends on σ², not σ. **Practical Rules for Assigning Weights** *Rule 1 — From standard deviations:* $$w_i \propto \frac{1}{\sigma_i^2}$$ If σ₁ = 0.01 m and σ₂ = 0.02 m, then w₁/w₂ = σ₂²/σ₁² = (0.02)²/(0.01)² = 4. So measurement 1 has 4× the weight of measurement 2. *Rule 2 — For differential leveling (by route distance):* $$w_i \propto \frac{1}{K_i}$$ where K_i is the length of the leveling route in km. Shorter routes accumulate fewer errors and thus carry higher weight. *Rule 3 — For repeated observations:* $$w_i \propto n_i$$ An observation that is itself the mean of n_i repeated measurements carries weight proportional to n_i (since σ_mean² = σ²/n, so 1/σ_mean² = n/σ² ∝ n). **Weighted Mean** When observations have different weights, the MPV is the **weighted mean** (also called the weighted arithmetic mean): $$\bar{x}_w = \frac{\sum w_i x_i}{\sum w_i} = \frac{w_1 x_1 + w_2 x_2 + \cdots + w_n x_n}{w_1 + w_2 + \cdots + w_n}$$ This is the least squares solution for the case of direct observations with different weights. It minimizes the weighted sum of squares: Σ(w_i v_i²) = minimum. **Standard Deviation of Unit Weight** In weighted adjustment, the standard deviation of unit weight (σ₀) is computed as: $$\sigma_0 = \sqrt{\frac{\sum w_i v_i^2}{n - 1}}$$ **Standard Deviation of the Weighted Mean:** $$\sigma_{\bar{x}_w} = \sigma_0 \sqrt{\frac{1}{\sum w_i}}$$
Examples
Observation 3 has the smallest σ (0.005 m), so it carries the highest weight (w₃' = 16) and dominates the result. Note that the weighted mean (250.138 m) is pulled toward x₃ = 250.139 m, which is the most precise measurement. Using relative weights that are whole numbers simplifies arithmetic without affecting the result.
Scenario
A distance is measured three times with the following results and standard deviations: x₁ = 250.132 m (σ₁ = 0.010 m), x₂ = 250.148 m (σ₂ = 0.020 m), x₃ = 250.139 m (σ₃ = 0.005 m). Find the weighted mean.
Solution
Step 1 — Compute weights (using c = 1): w₁ = 1/(0.010)² = 1/0.0001 = 10,000 w₂ = 1/(0.020)² = 1/0.0004 = 2,500 w₃ = 1/(0.005)² = 1/0.000025 = 40,000 Σw = 10,000 + 2,500 + 40,000 = 52,500 Step 2 — To simplify, use relative weights (divide all by 2,500): w₁' = 4, w₂' = 1, w₃' = 16 Σw' = 21 Step 3 — Compute weighted mean: x̄_w = (4×250.132 + 1×250.148 + 16×250.139) / 21 = (1000.528 + 250.148 + 4002.224) / 21 = 5252.900 / 21 = 250.138 m
Route 2 is the shortest (1 km) and therefore the most reliable, carrying the highest weight (w₂' = 12). The MPV of 12.352 m is notably pulled toward ΔH₂ = 12.356 m. This principle is fundamental to geodetic leveling network adjustments as practiced by NAMRIA.
Scenario
Three leveling routes connect benchmarks BM-A and BM-B. The observed elevation difference and route lengths are: Route 1: ΔH₁ = 12.342 m, K₁ = 3 km; Route 2: ΔH₂ = 12.356 m, K₂ = 1 km; Route 3: ΔH₃ = 12.348 m, K₃ = 4 km. Find the MPV of the elevation difference.
Solution
Step 1 — Weights for leveling (w ∝ 1/K): w₁ = 1/3, w₂ = 1/1 = 1, w₃ = 1/4 Step 2 — Use relative weights (multiply all by 12, the LCM of 3, 1, 4): w₁' = 4, w₂' = 12, w₃' = 3 Σw' = 19 Step 3 — Weighted mean: ΔH̄_w = (4×12.342 + 12×12.356 + 3×12.348) / 19 = (49.368 + 148.272 + 37.044) / 19 = 234.684 / 19 = 12.352 m
Applications
- Geodetic leveling network adjustment: Different routes have different weights based on distance.
- GPS baseline adjustment: Baselines are weighted by their variance-covariance matrices.
- Combined GPS/leveling: Different observation types (GNSS, optical leveling) get different weights.
- Cadastral surveys (PD 1529, CA 141): Weighted adjustment of traverse measurements.
- Triangulation/Trilateration: Angles and distances may have different weights based on instrument precision.
Misconceptions
- WRONG: 'Weight is proportional to 1/σ.' CORRECT: Weight is proportional to 1/σ². This is the most common board exam trap on this topic.
- WRONG: 'For leveling, weight ∝ distance.' CORRECT: Weight ∝ 1/distance (inverse). Longer route = more accumulated error = LOWER weight.
- WRONG: 'Changing the constant c changes the weighted mean.' CORRECT: The constant c cancels in the weighted mean formula. Using w = 1/σ² or w = 100/σ² gives the same x̄_w.
- WRONG: 'If one observation has twice the weight of another, it is twice as precise.' CORRECT: Twice the weight means variance is half, meaning σ is only 1/√2 ≈ 0.707 times smaller, not half.
Related Concepts
- Arithmetic Mean (special case of equal weights)
- Variance-Covariance Matrix in Least Squares
- Propagation of Weights
- Standard Deviation of Unit Weight
- Normal Equations in Least Squares
Common Exam Questions
Example
Q: Two observations, x₁ = 100.02 m (σ = 0.02 m) and x₂ = 100.05 m (σ = 0.01 m). Find weighted mean. A: w₁ = 2500, w₂ = 10000. x̄_w = (2500×100.02 + 10000×100.05)/12500 = 100.044 m.
Approach
Step 1: Compute w_i = 1/σ_i². Step 2: Simplify to relative weights if needed. Step 3: x̄_w = Σ(w_i x_i)/Σw_i.
Question Type
Computation — Weighted mean from σ values
Example
Q: Leveling routes of 2, 1, 4 km give elevations 25.342, 25.356, 25.348 m. Find MPV. A: w = 6, 12, 3 (relative, LCM=12). x̄_w = (6×25.342 + 12×25.356 + 3×25.348)/21.
Approach
Assign w_i = 1/K_i or use 1/K as relative weight. Then apply weighted mean formula.
Question Type
Computation — Weighted mean from leveling distances
Example
Q: Observation A has σ = 3 mm; Observation B has σ = 6 mm. What is the ratio of their weights (w_A : w_B)? A: w_A/w_B = σ_B²/σ_A² = 36/9 = 4. So w_A : w_B = 4 : 1.
Approach
Weight ratio = inverse of variance ratio = (σ_j/σ_i)² when comparing w_i to w_j.
Question Type
Conceptual — Ratio of weights
Key Points To Remember
- Weight w_i = c/σ_i². Weight is inversely proportional to VARIANCE (σ²), not to σ.
- For leveling: w_i ∝ 1/K_i (inverse of route distance in km). Shortest route = highest weight.
- For repeated observations: w_i ∝ n_i (number of repetitions). More reps = higher weight.
- Weighted mean: x̄_w = Σ(w_i × x_i) / Σw_i.
- If all weights are equal, the weighted mean reduces to the arithmetic mean — a good check.
- Observation with double the σ has ONE-QUARTER the weight (not half).
- Board exam trap: Do NOT divide weights by σ — divide by σ².
Measures of Precision and Accuracy
Before performing any adjustment, a geodetic engineer must evaluate the quality of measurements using numerical precision indicators. These measures also appear in survey specifications and acceptance criteria. **Precision vs. Accuracy** - **Precision** describes the consistency or reproducibility of measurements — how closely repeated measurements agree with each other. A precise measurement has small random errors. - **Accuracy** describes how close the measurements are to the true value — it accounts for both random AND systematic errors. **Key Precision Measures:** *1. Standard Deviation (σ)* Most widely used. Defines a range (x̄ ± σ) within which approximately 68.3% of observations fall. $$\sigma = \sqrt{\frac{\sum v_i^2}{n - 1}}$$ *2. Probable Error (PE)* Defines a range (x̄ ± PE) within which exactly 50% of observations fall. $$PE = 0.6745\sigma$$ The probable error of the mean: PE_m = 0.6745σ/√n *3. 90% Error (E₉₀)* $$E_{90} = 1.6449\sigma$$ *4. 95% Error (E₉₅) — 2σ rule* $$E_{95} \approx 1.960\sigma \approx 2\sigma$$ *5. 99% Error (E₉₉) — 3σ rule* $$E_{99} \approx 2.576\sigma \approx 3\sigma$$ These probability levels are critical in test criteria: if a residual exceeds 3σ, the observation is flagged as a potential blunder (outlier test). **Relative Precision (Relative Error)** For linear measurements, precision is often expressed as a ratio: $$\text{Relative Precision} = \frac{\sigma_{\bar{x}}}{\bar{x}} = \frac{1}{\bar{x}/\sigma_{\bar{x}}}$$ For example, if σ_x̄ = 0.005 m and x̄ = 500 m, then Relative Precision = 0.005/500 = 1/100,000. Survey standards specify minimum acceptable relative precision: - First-order traverse: 1:100,000 - Second-order: 1:50,000 - Third-order: 1:10,000
Examples
There is a 50% chance the true error is within ±0.008 m, a 95% chance it is within ±0.024 m, and a 99% chance within ±0.031 m. If a residual exceeds 3σ = 3 × 0.012 = 0.036 m, the measurement is suspect.
Scenario
The standard deviation of a single tape measurement is σ = 0.012 m. Find: (a) the probable error, (b) the 95% error, (c) the 99% error.
Solution
(a) PE = 0.6745 × 0.012 = 0.00809 m ≈ 0.008 m (b) E₉₅ = 1.960 × 0.012 = 0.02352 m ≈ 0.024 m (c) E₉₉ = 2.576 × 0.012 = 0.03091 m ≈ 0.031 m
Applications
- NAMRIA horizontal control: First-order networks require relative precision of 1:100,000.
- Cadastral survey acceptance: PRC standards specify maximum allowable closure errors.
- GNSS network design: Required accuracy determines the observation session duration.
- Building layout surveys: The 3σ rule is used to identify blunders in setout checks.
Misconceptions
- WRONG: 'Probable error = 0.6745 × mean.' CORRECT: PE = 0.6745 × σ (standard deviation, not mean).
- WRONG: 'A precise survey is always accurate.' CORRECT: High precision means low random error, but there may still be systematic errors making it inaccurate.
- WRONG: '1σ means 100% of errors fall within that range.' CORRECT: 1σ covers only 68.3% of the normal distribution.
Related Concepts
- Normal Distribution
- Confidence Intervals
- Outlier Detection
- Survey Order/Class Standards
- Error Propagation
Common Exam Questions
Example
Q: The probable error of a measurement is 0.020 m. Find σ. A: σ = PE/0.6745 = 0.020/0.6745 = 0.02965 m ≈ 0.030 m.
Approach
Use the multiplier relationships: PE = 0.6745σ; E₉₅ = 1.960σ; E₉₉ = 2.576σ. Solve algebraically for the unknown.
Question Type
Conversion between σ, PE, and E₉₅
Key Points To Remember
- Precision = consistency of repeat measurements (internal); Accuracy = closeness to truth (external).
- PE = 0.6745σ (50% probability range). Remember: 0.6745 is the z-score at 50th percentile.
- 68.3% of random errors fall within ±1σ; 95% within ±2σ; 99.7% within ±3σ.
- Residuals exceeding 3σ are flagged as potential blunders.
- Relative Precision = σ_x̄/x̄, expressed as 1/N (e.g., 1:100,000).
- Board exam: Given PE, find σ: σ = PE/0.6745. Given σ, find PE: PE = 0.6745σ.
Practice Problems
This is a classic board problem combining all the fundamental formulas. Key steps: work with the seconds part only to simplify arithmetic; always verify Σv = 0; use (n−1) = 4 in the denominator for σ; and report the final answer with its precision indicator. The probable error of 1.15'' means there is a 50% probability that the true angle lies within 35°14'20'' ± 1.15''.
Problem
Problem 1 (Board-Style): Five measurements of the same horizontal angle are recorded as follows: 35°14'22'', 35°14'18'', 35°14'25'', 35°14'20'', 35°14'15''. Find: (a) the Most Probable Value, (b) the residuals (verify Σv = 0), (c) the standard deviation of a single observation, (d) the standard deviation of the mean, and (e) the probable error of the mean.
Solution
Let the readings in seconds only be (adding 35°14' to each): 22, 18, 25, 20, 15. (a) MPV = Σx/n: Sum of seconds = 22 + 18 + 25 + 20 + 15 = 100'' Mean seconds = 100/5 = 20'' MPV = 35°14'20'' (b) Residuals (v_i = x_i − x̄): v₁ = 22 − 20 = +2'' v₂ = 18 − 20 = −2'' v₃ = 25 − 20 = +5'' v₄ = 20 − 20 = 0'' v₅ = 15 − 20 = −5'' Σv = 2 − 2 + 5 + 0 − 5 = 0'' ✓ (c) Σv² = 4 + 4 + 25 + 0 + 25 = 58 square-arcseconds σ = √(58/4) = √14.5 = 3.81'' (d) σ_x̄ = σ/√n = 3.81/√5 = 3.81/2.236 = 1.70'' (e) PE_mean = 0.6745 × σ_x̄ = 0.6745 × 1.70 = 1.15'' Final answer: MPV = 35°14'20'' ± 1.15'' (probable error)
Run 3 (1 km) is shortest and carries the highest weight (10), pulling the MPV toward ΔH₃ = 45.228 m. Run 2 (5 km) has the lowest weight (2) and least influence. The key insight: in leveling networks, a short, precise run always dominates the weighted mean. This principle applies directly to geodetic leveling network adjustments in Philippine national geodetic surveys.
Problem
Problem 2 (Board-Style): An elevation difference is determined by three independent leveling runs with the following results. Run 1: ΔH = 45.234 m, route length = 2 km. Run 2: ΔH = 45.241 m, route length = 5 km. Run 3: ΔH = 45.228 m, route length = 1 km. Determine the Most Probable Value of the elevation difference.
Solution
Step 1 — Assign weights (w ∝ 1/K): w₁ = 1/2, w₂ = 1/5, w₃ = 1/1 Step 2 — Convert to relative whole-number weights (multiply all by LCM = 10): w₁' = 5, w₂' = 2, w₃' = 10 Σw' = 17 Step 3 — Compute weighted mean: ΔH̄_w = (5×45.234 + 2×45.241 + 10×45.228) / 17 = (226.170 + 90.482 + 452.280) / 17 = 768.932 / 17 = 45.231 m Verification by residuals: v₁ = 45.234 − 45.231 = +0.003 m v₂ = 45.241 − 45.231 = +0.010 m v₃ = 45.228 − 45.231 = −0.003 m Check: Σ(w'v) = 5(0.003) + 2(0.010) + 10(−0.003) = 0.015 + 0.020 − 0.030 = 0.005 ≈ 0 (small rounding) ✓
The ratio w₁:w₂ = 1:4 because σ₂ is half of σ₁, making w₂ = (σ₁/σ₂)² = (2)² = 4 times larger. Adding measurement 3 (σ₃ = 0.040 m, the least precise) barely changes the result from 875.447 m because its weight is only 1 compared to 16 for the most precise. This demonstrates how a very precise observation dominates the weighted mean.
Problem
Problem 3 (Board-Style): Two measurements of a slope distance are made: x₁ = 875.432 m with σ₁ = 0.020 m, and x₂ = 875.451 m with σ₂ = 0.010 m. (a) Find the weighted mean. (b) What is the ratio w₁:w₂? (c) If a third measurement x₃ = 875.440 m with σ₃ = 0.040 m is added, find the new weighted mean.
Solution
(a) Two-measurement case: w₁ = 1/(0.020)² = 2,500 w₂ = 1/(0.010)² = 10,000 Relative weights: w₁' = 1, w₂' = 4 (divide by 2500) Σw' = 5 x̄_w = (1×875.432 + 4×875.451)/5 = (875.432 + 3501.804)/5 = 4377.236/5 = 875.447 m (b) w₁:w₂ = 2500:10000 = 1:4 (or equivalently, σ₂²:σ₁² = 0.0001:0.0004 = 1:4) (c) Three-measurement case: w₃ = 1/(0.040)² = 625 Relative weights (divide all by 625): w₁'' = 2500/625 = 4 w₂'' = 10000/625 = 16 w₃'' = 625/625 = 1 Σw'' = 21 x̄_w = (4×875.432 + 16×875.451 + 1×875.440)/21 = (3501.728 + 14007.216 + 875.440)/21 = 18384.384/21 = 875.447 m (rounded to mm)
When the only difference between observations is the number of repetitions (and all single measurements have the same precision), weight is directly proportional to n. Team B's result (9 measurements) has the most influence. This is because the mean of 9 measurements has σ_mean = σ/√9 = σ/3, meaning it is 3× more precise than a single measurement, and 9× the variance-based weight of Team C.
Problem
Problem 4 (Board-Style): A distance is measured by three survey teams using different methods and equipment. Team A: 1,250.345 m, measured 4 times. Team B: 1,250.372 m, measured 9 times. Team C: 1,250.358 m, measured 1 time. Find the Most Probable Value assuming equal per-measurement precision.
Solution
Since all measurements have equal per-measurement precision, weights are proportional to the number of repetitions: w_A = 4, w_B = 9, w_C = 1 Σw = 14 x̄_w = (4×1250.345 + 9×1250.372 + 1×1250.358) / 14 = (5001.380 + 11253.348 + 1250.358) / 14 = 17505.086 / 14 = 1250.363 m
Note that PE of mean = PE_single/√n = 0.018/4 = 0.0045 m. This clean result confirms: the probable error of the mean improves as √n. Also note: the 95% bounds use 1.960σ (not 2σ exactly — though 2σ is a common approximation). The PE is exactly half of the '50% interval' — a useful mnemonic: PE corresponds to 50% probability.
Problem
Problem 5 (Board-Style — Probable Error Application): A series of distance measurements has a probable error of PE = 0.018 m for a single measurement. (a) What is the standard deviation of a single measurement? (b) What is the probable error of the mean of 16 measurements? (c) What is the standard deviation of the mean of 16 measurements? (d) Between what bounds will 95% of single measurements fall?
Solution
(a) σ = PE/0.6745 = 0.018/0.6745 = 0.02669 m ≈ 0.027 m (b) PE_mean = PE/√n = 0.018/√16 = 0.018/4 = 0.0045 m (c) σ_mean = σ/√n = 0.02669/4 = 0.006672 m ≈ 0.0067 m (Verification: PE_mean = 0.6745 × 0.0067 = 0.00452 ≈ 0.0045 ✓) (d) E₉₅ = 1.960σ = 1.960 × 0.02669 = 0.05231 m ≈ 0.052 m Bounds: x̄ ± 0.052 m (i.e., 95% of measurements fall within 0.052 m of the mean)
Exam Preparation Tips
- MASTER THE WEIGHT FORMULA: The single most common board exam trap is using w = 1/σ instead of w = 1/σ². Drill this until it is automatic: weight is inversely proportional to VARIANCE, not standard deviation.
- MEMORIZE THE KEY MULTIPLIERS: PE = 0.6745σ; E₉₅ = 1.960σ; E₉₉ = 2.576σ; σ_mean = σ/√n. These appear repeatedly across board exam cycles.
- PRACTICE THE FULL WORKFLOW: For any MPV problem, always follow this sequence: (1) Compute mean, (2) Compute residuals, (3) Verify Σv = 0, (4) Compute Σv², (5) Apply σ formula with (n−1). A Σv ≠ 0 check catches arithmetic errors before they propagate.
- LEVELING WEIGHT RULE: For differential leveling problems, w ∝ 1/K. Commit to memory: shorter route = smaller K = larger weight = more reliable. This is the standard rule in geodetic leveling network adjustments.
- RELATIVE WEIGHTS SIMPLIFY COMPUTATION: When weights are fractions (e.g., 1/2, 1/3, 1/5), multiply all by the LCM to get whole-number relative weights. The weighted mean is identical — just easier to compute.
- KNOW THE ERROR CLASSIFICATION FOR FILL-IN AND MULTIPLE CHOICE: Blunders → eliminate; Systematic → correct; Random → adjust. This three-part answer is one of the most tested conceptual questions.
- BESSEL'S CORRECTION IS MANDATORY: Always use (n−1) in the denominator for σ of a sample. The exam will offer both n and (n−1) as choices. Sample standard deviation always uses (n−1).
- USE DIMENSIONAL ANALYSIS: If weights from σ values seem unreasonably large (e.g., 10,000), do not worry — they cancel in the weighted mean formula. What matters is the RATIO of weights.
- UNDERSTAND WHY PRECISION IMPROVES AS √n, NOT n: Explain this to yourself: σ_mean = σ/√n. To halve σ_mean, you need 4× observations. To reduce by 10×, you need 100× observations. This concept is tested both computationally and conceptually.
- CONNECT TO LEAST SQUARES PRINCIPLE: The arithmetic mean minimizes Σv² for equal weights; the weighted mean minimizes Σ(wv²) for unequal weights. Knowing this connection — not just the formulas — shows mastery expected at the licensure level.
- REVIEW PHILIPPINE SURVEY STANDARDS: NAMRIA/DPWH geodetic control specifications define precision requirements in terms of relative precision (1:100,000 for first-order). Know how to compute and compare relative precision to these benchmarks.
- PRACTICE UNDER TIME PRESSURE: Board exam problems on this topic are computation-heavy. Aim to solve a standard MPV-plus-σ-plus-σ_mean problem in under 4 minutes. Use relative weights and keep arithmetic organized in a table format.
In summary
The Theory of Errors, Weights, and Most Probable Value is not merely an academic exercise — it is the quantitative foundation upon which every reliable geodetic product rests, from cadastral titles under PD 1529 to the national geodetic reference frame (PRS92/WGS84) maintained by NAMRIA. As a geodetic engineer, you will encounter these concepts in every project: choosing how many measurements to take, deciding which instrument to use, flagging suspicious observations, and computing the best estimate of a quantity from redundant data. For the PRC licensure examination, master these four pillars: (1) Classify errors correctly — eliminate blunders, correct systematic errors, adjust random errors. (2) Compute the arithmetic mean and its standard deviation using σ = √(Σv²/(n−1)) and σ_x̄ = σ/√n — always verify Σv = 0. (3) Assign weights correctly — w = 1/σ² (not 1/σ), w ∝ 1/K for leveling, w ∝ n for repetitions. (4) Compute the weighted mean as x̄_w = Σ(wx)/Σw. These fundamentals seamlessly connect to the more advanced topics that follow in this course: propagation of errors, least squares adjustment of traverses, level networks, and GNSS networks. Every normal equation in a least squares adjustment, every variance-covariance matrix in GPS processing, and every blunder detection statistic traces back to exactly these principles. Build your foundation on these concepts solidly, and the rest of adjustment computations will follow naturally. Magtrabaho ng husay — the board exam rewards engineers who understand not just the formulas, but the reasoning behind them.
Ready to practise for the GELE 2026?
Super Tutor's AI review plan adapts to your weak areas and builds a weekly practice schedule around your target GELE exam date.